AMD Radeon AI PRO 9600D vs NVIDIA GB10 Comparison

Head-to-Head Benchmarks

The NVIDIA GB10 holds the only recorded benchmark scores in this comparison. The database shows two Geekbench results for the GB10: an OpenCL score of 120,137 and a Vulkan score of 114,648. These produce an average benchmark score of 117,393, placing the GB10 in the 95th percentile among all GPUs tracked by the database. The AMD Radeon AI PRO 9600D has no recorded benchmark entries, leaving its performance profile unquantified in direct comparison. The GB10 sits 0.3% ahead of the NVIDIA RTX 4000 SFF Ada Generation, which averages 117,088 points. Against the AMD Radeon PRO W7700, the GB10 trails by 1.3%, as that card posts an average score of 118,976. The gap widens favorably for the GB10 against the NVIDIA Tesla V100 SXM2 16 GB, where it leads by 2.6% (114,395 average), and against the NVIDIA RTX A5500 Mobile, where it leads by 3% (113,944 average).

The raw compute figures favor the GB10 in several key metrics. The GB10 delivers 29.71 TFLOPS of FP32 performance compared to 24.82 TFLOPS for the AMD part, a clear advantage of roughly 4.89 TFLOPS. Both cards match their FP32 output in FP16 with a 1:1 ratio, so the same margin applies in half-precision workloads. The texture rate also leans heavily toward NVIDIA: the GB10 reaches 928.5 GTexel/s, while the AMD card manages 387.8 GTexel/s, more than double the throughput. However, the AMD Radeon AI PRO 9600D wins decisively in pixel throughput, posting 193.9 GPixel/s against 116.1 GPixel/s for the GB10, a margin of roughly 67%.

Memory capacity strongly favors the GB10 with 128 GB of LPDDR5X, quadruple the 32 GB of GDDR6 found on the AMD card. Yet the AMD memory subsystem delivers far higher bandwidth at 576.0 GB/s versus 273.2 GB/s for the GB10. The GB10 compensates with a higher memory clock in terms of base frequency: 1067 MHz with 8.5 Gbps effective transfer, while the AMD card runs at 2250 MHz with 18 Gbps effective. The bus width is identical at 256 bits for both.

Clock speeds show the NVIDIA part running faster in both base and boost states. The GB10 has a base clock of 1665 MHz and a boost clock of 2418 MHz, compared to 1080 MHz base and 2020 MHz boost for the AMD Radeon AI PRO 9600D. The GB10 also packs double the shading units (6144 versus 3072), double the texture mapping units (384 versus 192), and eight times the tensor cores (384 versus none listed for AMD). Both cards feature 48 ray tracing cores and 96 ROPs for AMD versus 48 ROPs for NVIDIA, the latter being half the count.

Where Each One Wins

The AMD Radeon AI PRO 9600D wins in scenarios that depend on pixel generation and memory bandwidth. Its 193.9 GPixel/s fill rate, built on 96 ROPs, gives it a strong edge in rasterization-heavy tasks where pixel output is the limiting factor. The 576.0 GB/s memory bandwidth, paired with 18 Gbps effective GDDR6, supports high-resolution texture streaming and large framebuffer workloads. The AMD card also consumes slightly more power at 150 W TDP versus 140 W for the GB10, but that extra power buys a higher pixel rate.

The NVIDIA GB10 dominates in compute-oriented workloads. Its 29.71 TFLOPS FP32 and FP16 output, combined with 384 tensor cores, positions it for AI inference, machine learning training, and scientific simulation. The 928.5 GTexel/s texture rate indicates strong geometry and texture processing capability. The 128 GB memory capacity allows massive datasets to reside on-card without host memory transfers, a clear advantage for large model loading and high-performance computing applications. The GB10 also carries a higher average benchmark score in the database, reinforcing its measured performance lead over the AMD part in available tests.

The AMD card wins on API compatibility for graphics. It supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, whereas the GB10 lists N/A for all three graphics APIs. This makes the AMD Radeon AI PRO 9600D the only one of the two suitable for traditional gaming or DirectX-based rendering workloads. The AMD card also offers a DisplayPort 2.1a output, while the GB10 provides a single HDMI connection.

Architecture Differences

The two GPUs come from different architectural generations and process nodes. The AMD Radeon AI PRO 9600D uses the Navi 48 chip built on RDNA 4.0 architecture, manufactured on a 4 nm TSMC process. It belongs to the Radeon Pro Navi (Navi IV Series) generation. The die measures 357 mm² and contains 53,900 million transistors, yielding a density of 151.0M transistors per mm². The GB10 uses the GB20B chip based on Blackwell 2.0 architecture, fabricated on a 5 nm TSMC process. Its die is slightly larger at 382 mm², but transistor count is listed as unknown in the database. The GB10 belongs to the Server Blackwell (Bxx) generation, with its predecessor being Server Hopper and its successor being Server Rubin.

Memory technologies differ fundamentally. AMD pairs its 32 GB pool with GDDR6 memory, while NVIDIA uses 128 GB of LPDDR5X. Both operate over a 256-bit bus, but the AMD memory runs at 2250 MHz with 18 Gbps effective transfer, producing 576.0 GB/s bandwidth. The NVIDIA memory runs at 1067 MHz with 8.5 Gbps effective transfer, producing 273.2 GB/s bandwidth. The GB10's lower clock speed is offset by its much larger capacity, suggesting a design optimized for holding large working sets rather than maximizing streaming throughput.

Core counts diverge sharply. The GB10 has 6144 shading units, 384 TMUs, 48 ROPs, 48 RT cores, and 384 tensor cores. The AMD card has 3072 shading units, 192 TMUs, 96 ROPs, 48 RT cores, and no tensor cores listed in the database. This configuration gives NVIDIA a 2:1 advantage in shader and texture hardware but leaves AMD with a 2:1 advantage in ROP count. The presence of tensor cores on the GB10 and their absence on the AMD card marks a clear functional split: the GB10 targets tensor-accelerated workloads, while the AMD card relies on conventional shader and raster pipelines.

Physical and power characteristics also differ. The AMD card is a single-slot design measuring 241 mm in length, 111 mm in height, and 19 mm in width, requiring a 16-pin power connector and a 450 W suggested PSU. The GB10 is an IGP (integrated graphics processor) form factor measuring 150 mm by 51 mm by 150 mm, with no power connectors and a 300 W suggested PSU. Both use PCIe 5.0 x16 interfaces. The AMD card consumes 150 W TDP; the GB10 consumes 140 W. Release dates place the GB10 first at 2025-10-14, followed by the AMD card at 2025-12-10.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GB10 has an average benchmark score of 117,393, placing it in the 95th percentile. The AMD Radeon AI PRO 9600D has no recorded benchmark scores, so it sits at the 50th percentile with an average score of 0.

Q: How does the GB10 compare to its nearest rivals in the database?

A: The GB10 is 0.3% ahead of the NVIDIA RTX 4000 SFF Ada Generation (117,088), 1.3% behind the AMD Radeon PRO W7700 (118,976), 2.6% ahead of the NVIDIA Tesla V100 SXM2 16 GB (114,395), and 3% ahead of the NVIDIA RTX A5500 Mobile (113,944).

Q: What is the memory configuration difference between the two cards?

A: The AMD Radeon AI PRO 9600D has 32 GB of GDDR6 with a 256-bit bus and 576.0 GB/s bandwidth. The NVIDIA GB10 has 128 GB of LPDDR5X with a 256-bit bus and 273.2 GB/s bandwidth.

Q: Which card supports DirectX, OpenGL, and Vulkan?

A: Only the AMD Radeon AI PRO 9600D supports these APIs: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The NVIDIA GB10 lists N/A for all three graphics APIs.

Q: What are the core counts for each GPU?

A: The AMD card has 3072 shading units, 192 TMUs, 96 ROPs, and 48 RT cores. The GB10 has 6144 shading units, 384 TMUs, 48 ROPs, 48 RT cores, and 384 tensor cores.

Q: What is the launch MSRP for the NVIDIA GB10?

A: The NVIDIA GB10 has a launch MSRP of 3,999 USD. The AMD Radeon AI PRO 9600D has no launch MSRP listed in the database.

The Verdict

The data separates these two GPUs into distinct use cases. The NVIDIA GB10 delivers measured performance, with its 117,393 average benchmark score and 95th percentile ranking confirming strong real-world results. Its 29.71 TFLOPS FP32 output, 384 tensor cores, and 128 GB memory capacity make it the clear choice for compute-intensive workloads such as AI inference, machine learning, and large-scale data processing. The GB10 also leads in texture throughput at 928.5 GTexel/s and carries double the shading units of the AMD part. Its 140 W TDP and IGP form factor with no power connectors indicate a design intended for dense server integration.

The AMD Radeon AI PRO 9600D offers advantages in specific areas despite lacking benchmark data. Its 193.9 GPixel/s pixel rate, built on 96 ROPs, makes it better suited for rasterization-heavy graphics workloads. The 576.0 GB/s memory bandwidth exceeds the GB10 by more than double, supporting high-bandwidth streaming applications. The AMD card also supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, making it the only option here for standard graphics APIs. Its 4 nm process node gives it a transistor density of 151.0M per mm², though the GB10's 5 nm process with unknown transistor count does not allow a direct density comparison.

For buyers needing measured compute performance and massive memory capacity, the GB10 is the only choice with recorded benchmark results. For workloads requiring pixel fill rates, high memory bandwidth, or graphics API compatibility, the AMD Radeon AI PRO 9600D is the better fit. The absence of benchmark scores for the AMD card means its real-world performance cannot be verified against the GB10's recorded results. The 95th percentile ranking of the GB10, alongside its leads over the RTX 4000 SFF Ada Generation and RTX A5500 Mobile, confirms it as the higher-performing part in the database's measurements.